34
1 320 1 011 APC9 2011 York

Mon 10.00 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

Embed Size (px)

Citation preview

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 139

1

983107983151983149983152983137983154983137983156983145983158983141 983105983150983137983148983161983155983145983155 983151983142 983107983151983150983156983154983151983148983148983141983154

983124983141983139983144983150983151983148983151983143983145983141983155 983142983151983154 983137 983107983112983119320 983110983141983140 983106983137983156983139983144

983110983141983154983149983141983150983156983137983156983145983151983150

983123983156983141983152983144983141983150 983107983154983137983158983141983150

1983097983156983144 983123983141983152983156983141983149983138983141983154 983090011

APC9 2011

York

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 239

22

983120983154983141983155983141983150983156983137983156983145983151983150 983119983158983141983154983158983145983141983159

983085 983119983158983141983154983158983145983141983159 983151983142 983120983105983124983085983106983145983151 983120983154983151983146983141983139983156

983085 983107983151983150983156983154983151983148983148983141983154 983123983145983149983157983148983137983156983145983151983150983155

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983117983120983107 983107983151983150983156983154983151983148

983085 983109983160983152983141983154983145983149983141983150983156983137983148 983126983141983154983145983142983145983139983137983156983145983151983150983085 983107983151983150983139983148983157983155983145983151983150983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 339

3

983120983105983124983085983106983145983151 983120983154983151983146983141983139983156

Industry ndashled research program funded by EnterpriseIreland which aims to develop a PAT enabled mammalian

cell bioprocess

Modelling andPrediction

ProcessControl

Online

Sampling

AnalyticsSensors

Real timemonitoring

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 439

4

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148983148983145983150983143 983137983150983140 983107983151983150983156983154983151983148

983107983144983137983148983148983141983150983143983141983155 983151983142 983138983145983151983154983141983137983139983156983151983154 983139983151983150983156983154983151983148

983085 983124983144983141983145983154 983140983161983150983137983149983145983139 983138983141983144983137983158983145983151983157983154 983145983155 983145983150983144983141983154983141983150983156983148983161983150983151983150983148983145983150983141983137983154

983085 983105983139983139983157983154983137983156983141 983152983154983151983139983141983155983155 983149983151983140983141983148983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983140983157983141 983156983151 983156983144983141 983139983151983149983152983148983141983160983145983156983161 983151983142 983156983144983141983157983150983140983141983154983148983161983145983150983143 983138983145983151983139983144983141983149983145983139983137983148 983152983154983151983139983141983155983155983141983155

983085 983117983151983140983141983148 983152983137983154983137983149983141983156983141983154983155 983158983137983154983161 983145983150 983137983150

983157983150983152983154983141983140983145983139983156983137983138983148983141 983149983137983150983150983141983154

983085 983122983141983148983145983137983138983148983141 983138983145983151983155983141983150983155983151983154983155 983156983151 983149983141983137983155983157983154983141983145983150983156983141983154983139983141983148983148983157983148983137983154 983137983139983156983145983158983145983156983145983141983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983149983137983147983145983150983143 983156983144983141 983152983154983151983139983141983155983155 983155983156983137983156983141983155

983158983141983154983161 983140983145983142983142983145983139983157983148983156 983156983151 983139983144983137983154983137983139983156983141983154983145983155983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 539

5

983120983105983124 983109983150983137983138983148983141983140 983110983141983140983085983138983137983156983139983144 983106983145983151983152983154983151983139983141983155983155

X

L

G

V

S in

F

X Biomass

G GlucoseL LactateV Volume

F Feed RateS in Inlet Substrate

Concentration

Process

Controller

Raman Probe

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 239

22

983120983154983141983155983141983150983156983137983156983145983151983150 983119983158983141983154983158983145983141983159

983085 983119983158983141983154983158983145983141983159 983151983142 983120983105983124983085983106983145983151 983120983154983151983146983141983139983156

983085 983107983151983150983156983154983151983148983148983141983154 983123983145983149983157983148983137983156983145983151983150983155

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983117983120983107 983107983151983150983156983154983151983148

983085 983109983160983152983141983154983145983149983141983150983156983137983148 983126983141983154983145983142983145983139983137983156983145983151983150983085 983107983151983150983139983148983157983155983145983151983150983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 339

3

983120983105983124983085983106983145983151 983120983154983151983146983141983139983156

Industry ndashled research program funded by EnterpriseIreland which aims to develop a PAT enabled mammalian

cell bioprocess

Modelling andPrediction

ProcessControl

Online

Sampling

AnalyticsSensors

Real timemonitoring

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 439

4

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148983148983145983150983143 983137983150983140 983107983151983150983156983154983151983148

983107983144983137983148983148983141983150983143983141983155 983151983142 983138983145983151983154983141983137983139983156983151983154 983139983151983150983156983154983151983148

983085 983124983144983141983145983154 983140983161983150983137983149983145983139 983138983141983144983137983158983145983151983157983154 983145983155 983145983150983144983141983154983141983150983156983148983161983150983151983150983148983145983150983141983137983154

983085 983105983139983139983157983154983137983156983141 983152983154983151983139983141983155983155 983149983151983140983141983148983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983140983157983141 983156983151 983156983144983141 983139983151983149983152983148983141983160983145983156983161 983151983142 983156983144983141983157983150983140983141983154983148983161983145983150983143 983138983145983151983139983144983141983149983145983139983137983148 983152983154983151983139983141983155983155983141983155

983085 983117983151983140983141983148 983152983137983154983137983149983141983156983141983154983155 983158983137983154983161 983145983150 983137983150

983157983150983152983154983141983140983145983139983156983137983138983148983141 983149983137983150983150983141983154

983085 983122983141983148983145983137983138983148983141 983138983145983151983155983141983150983155983151983154983155 983156983151 983149983141983137983155983157983154983141983145983150983156983141983154983139983141983148983148983157983148983137983154 983137983139983156983145983158983145983156983145983141983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983149983137983147983145983150983143 983156983144983141 983152983154983151983139983141983155983155 983155983156983137983156983141983155

983158983141983154983161 983140983145983142983142983145983139983157983148983156 983156983151 983139983144983137983154983137983139983156983141983154983145983155983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 539

5

983120983105983124 983109983150983137983138983148983141983140 983110983141983140983085983138983137983156983139983144 983106983145983151983152983154983151983139983141983155983155

X

L

G

V

S in

F

X Biomass

G GlucoseL LactateV Volume

F Feed RateS in Inlet Substrate

Concentration

Process

Controller

Raman Probe

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 339

3

983120983105983124983085983106983145983151 983120983154983151983146983141983139983156

Industry ndashled research program funded by EnterpriseIreland which aims to develop a PAT enabled mammalian

cell bioprocess

Modelling andPrediction

ProcessControl

Online

Sampling

AnalyticsSensors

Real timemonitoring

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 439

4

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148983148983145983150983143 983137983150983140 983107983151983150983156983154983151983148

983107983144983137983148983148983141983150983143983141983155 983151983142 983138983145983151983154983141983137983139983156983151983154 983139983151983150983156983154983151983148

983085 983124983144983141983145983154 983140983161983150983137983149983145983139 983138983141983144983137983158983145983151983157983154 983145983155 983145983150983144983141983154983141983150983156983148983161983150983151983150983148983145983150983141983137983154

983085 983105983139983139983157983154983137983156983141 983152983154983151983139983141983155983155 983149983151983140983141983148983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983140983157983141 983156983151 983156983144983141 983139983151983149983152983148983141983160983145983156983161 983151983142 983156983144983141983157983150983140983141983154983148983161983145983150983143 983138983145983151983139983144983141983149983145983139983137983148 983152983154983151983139983141983155983155983141983155

983085 983117983151983140983141983148 983152983137983154983137983149983141983156983141983154983155 983158983137983154983161 983145983150 983137983150

983157983150983152983154983141983140983145983139983156983137983138983148983141 983149983137983150983150983141983154

983085 983122983141983148983145983137983138983148983141 983138983145983151983155983141983150983155983151983154983155 983156983151 983149983141983137983155983157983154983141983145983150983156983141983154983139983141983148983148983157983148983137983154 983137983139983156983145983158983145983156983145983141983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983149983137983147983145983150983143 983156983144983141 983152983154983151983139983141983155983155 983155983156983137983156983141983155

983158983141983154983161 983140983145983142983142983145983139983157983148983156 983156983151 983139983144983137983154983137983139983156983141983154983145983155983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 539

5

983120983105983124 983109983150983137983138983148983141983140 983110983141983140983085983138983137983156983139983144 983106983145983151983152983154983151983139983141983155983155

X

L

G

V

S in

F

X Biomass

G GlucoseL LactateV Volume

F Feed RateS in Inlet Substrate

Concentration

Process

Controller

Raman Probe

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 439

4

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148983148983145983150983143 983137983150983140 983107983151983150983156983154983151983148

983107983144983137983148983148983141983150983143983141983155 983151983142 983138983145983151983154983141983137983139983156983151983154 983139983151983150983156983154983151983148

983085 983124983144983141983145983154 983140983161983150983137983149983145983139 983138983141983144983137983158983145983151983157983154 983145983155 983145983150983144983141983154983141983150983156983148983161983150983151983150983148983145983150983141983137983154

983085 983105983139983139983157983154983137983156983141 983152983154983151983139983141983155983155 983149983151983140983141983148983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983140983157983141 983156983151 983156983144983141 983139983151983149983152983148983141983160983145983156983161 983151983142 983156983144983141983157983150983140983141983154983148983161983145983150983143 983138983145983151983139983144983141983149983145983139983137983148 983152983154983151983139983141983155983155983141983155

983085 983117983151983140983141983148 983152983137983154983137983149983141983156983141983154983155 983158983137983154983161 983145983150 983137983150

983157983150983152983154983141983140983145983139983156983137983138983148983141 983149983137983150983150983141983154

983085 983122983141983148983145983137983138983148983141 983138983145983151983155983141983150983155983151983154983155 983156983151 983149983141983137983155983157983154983141983145983150983156983141983154983139983141983148983148983157983148983137983154 983137983139983156983145983158983145983156983145983141983155 983137983154983141 983154983137983154983141983148983161983137983158983137983145983148983137983138983148983141 983149983137983147983145983150983143 983156983144983141 983152983154983151983139983141983155983155 983155983156983137983156983141983155

983158983141983154983161 983140983145983142983142983145983139983157983148983156 983156983151 983139983144983137983154983137983139983156983141983154983145983155983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 539

5

983120983105983124 983109983150983137983138983148983141983140 983110983141983140983085983138983137983156983139983144 983106983145983151983152983154983151983139983141983155983155

X

L

G

V

S in

F

X Biomass

G GlucoseL LactateV Volume

F Feed RateS in Inlet Substrate

Concentration

Process

Controller

Raman Probe

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 539

5

983120983105983124 983109983150983137983138983148983141983140 983110983141983140983085983138983137983156983139983144 983106983145983151983152983154983151983139983141983155983155

X

L

G

V

S in

F

X Biomass

G GlucoseL LactateV Volume

F Feed RateS in Inlet Substrate

Concentration

Process

Controller

Raman Probe

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 639

6

983106983145983151983152983154983151983139983141983155983155 983117983151983140983141983148

983125983150983155983156983154983157983139983156983157983154983141983140 983150983151983150983085983155983141983143983154983141983143983137983156983141983140 983138983145983151983152983154983151983139983141983155983155 983149983151983140983141983148

983085 983123983161983155983156983141983149 983151983142 983139983151983157983152983148983141983140 983119983108983109983155 983149983137983155983155 983138983137983148983137983150983139983141 983137983150983140 983154983137983156983141 983141983153983157983137983156983145983151983150983155

983085 983107983151983150983156983137983145983150983155 983137983152983152983154983151983160983145983149983137983156983141983148983161 16 983149983151983140983141983148 983152983137983154983137983149983141983156983141983154983155

983085 983125983155983141983140 983137 983140983145983142983142983141983154983141983150983156983145983137983148 983141983158983151983148983157983156983145983151983150 983152983137983154983137983149983141983156983141983154 983151983152983156983145983149983145983155983137983156983145983151983150983137983148983143983151983154983145983156983144983149 983156983151 983142983145983156 983156983144983141 983152983137983154983137983149983141983156983141983154 983158983137983148983157983141983155 983159983145983156983144 983156983144983141 983152983151983155983155983145983138983148983141 983154983137983150983143983141983155

983139983151983150983155983156983154983137983145983150983141983140 983138983137983155983141983140 983151983150 983152983157983138983148983145983155983144983141983140 983158983137983148983157983141983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 739

7

983120983154983151983139983141983155983155 983107983151983150983156983154983151983148

ProcessObjective

MaximizeBiomassProductivity

ControlObjective

Maintain a lowGlucoseGlutamine level

in order to avoid a buildup of LactateAmmonia

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 839

8

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983140983141983158983141983148983151983152983149983141983150983156

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 939

9

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

bull Based on the ISA control law consisting of three control

parameters (K Td Ti)

bull Control parameters tuned via Ziegler Nichols (Closedloop) relay based tuning

Classical Control PID Controller

++= int

t

d

i dt

t de

T d eT t eK t u0

)(

)(

1

)()( τ τ

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1039

10

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Setpoint OutputProcess

P

I

D

)(t Ke

int t

i

d eT

K

0)( τ τ

dt

t deKT d

)(

sum sum+

-

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1139

11

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Classical Control PID Controller Tuning

Ziegler Nichols closed loop with relay identification technique

71

U K PID K =

T i =

T d

=

2

U P

8

U P

= 0010558

= 04

= 01

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1239

12

Control Strategy Development

Output

Setpoint

Process

ModelObjectiveFunction

Optimisation

Feed

Advanced Control MPC Controller

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1339

13

Control Strategy Development

Industrial MPC Formulation

sumsum =

minus++

=

+

∆∆

Λ∆+minusΓ minus+

M

i

ik ik

P

i

ik t ut u

t ut Rt y M k k 1

2

)1

2

1)()(()()((min

1

Subject to

maxmin

maxmin

uuu

y y y

∆le∆le∆

lele

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1439

14

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156

Advanced Control MPC Tuning

4 Tuning parameters

P = Prediction horizon

M = Control horizon

Γ = Output weight matrix

Λ = Input weight matrix

Offline Tuning Rules

P = Increase until no further effect

M = 1 (gt1 leads to more aggressive control)

Γ = scale on control variables = 1

Λ = Increase to avoid aggressive CO

Tuning results

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1539

15

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983155983145983149983157983148983137983156983145983151983150

983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1639

16

983107983151983150983156983154983151983148983148983141983154 983120983141983154983142983151983154983149983137983150983139983141 983123983145983149983157983148983137983156983145983151983150 983122983141983155983157983148983156983155

bullSet-point Tracking

bullDisturbance Compensation

bullPropagation of Measurement Noise

Performance Off-Line Testing Criteria

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1739

17

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1839

18

983120983113983108 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 1939

19

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2039

20

983117983120983107 983123983141983156983085983152983151983145983150983156 983123983156983141983152 983124983154983137983139983147983145983150983143

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2139

21

PID Disturbance Compensation

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M ) Cglc

Set-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2239

22

MPC Disturbance Compensation

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M

)

CglcSet-point

0

002

004

006

008

01

012

0 50 100 150 200 250 300Time (h)

F e e d r a

t e ( L h )

Controller Output

Disturbanceapplied

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2339

23

983120983113983108 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

Fast lsquoon-offrsquo

Control

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2439

24

983120983113 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c

e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h )

Controller Output

Removal oflsquoDrsquo term leadsto less abruptcontrol

Better set-point tracking

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2539

25

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2639

26

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

510

15

20

25

30

0 50 100 150 200 250 300Time (h)

C o n c e n t r a t i o n ( m M ) Cglc

Set-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h )

Controller Output

P = 10

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2739

27

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M

)

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2839

28

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c

e n t r a t i o n ( m M )

CglcSet-point

0001002003004005006007

0 50 100 150 200 250 300

Time (h)

F e e d - r a t e ( L h )

Controller Output

P = 40

M = 1

Γ = 1

Λ = 0

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 2939

29

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a

t e ( L h

Controller Output

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3039

30

983117983120983107 983120983154983151983152983137983143983137983156983145983151983150 983151983142 983117983141983137983155983157983154983141983149983141983150983156 983118983151983145983155983141

0

5

10

15

20

25

30

0 50 100 150 200 250 300

Time (h)

C o n c e

n t r a t i o n ( m M )

CglcSet-point

0

001

002

003

004

005

006

007

0 50 100 150 200 250 300Time (h)

F e e d - r a t e ( L h

Controller Output

P = 40

M = 1

Γ = 1

Λ = 10000

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3139

31

983123983145983149983157983148983137983156983145983151983150 983123983156983157983140983161 983122983141983155983157983148983156983155

bull Reference tracking MPC starts adjusting the control signalahead of reference changes while PID cannot start beforeMPC gives less control error

bull Disturbance compensation MPC better than PID MPCcan deal with plant-model mismatch

bull Measurement noise Less propagation through tuned MPCthan tuned PID By increasing the penalty weight on the input

the control action in the presence of measurement noise canbe dampened to a more smooth control signal

bullSISOMIMO PID can only handle one input and one output

MPC is a multivariable control strategy

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3239

32

983107983151983150983156983154983151983148 983123983156983154983137983156983141983143983161 983108983141983158983141983148983151983152983149983141983150983156 983123983145983149983157983148983137983156983145983151983150 amp 983105983152983152983148983145983139983137983156983145983151983150

983107983151983150983156983154983151983148 983155983156983154983137983156983141983143983161 983154983141983137983148 983156983145983149983141 983137983152983152983148983145983139983137983156983145983151983150983085 983120983113983108 983107983151983150983156983154983151983148

983085 983118983151983150983085983148983145983150983141983137983154 983117983120983107 (983117983151983140983141983148 983120983154983141983140983145983139983156983145983158983141 983107983151983150983156983154983151983148)

983107983151983150983142983145983143983157983154983141

amp 983124983157983150983141

983123983145983149983157983148983137983156983141

983122983141983137983148

983124983145983149983141

983148

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3339

33

3 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983148983148 983105 983148983145 983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3439

34

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983110983141983141983140 983120983157983149983152

983125983123983106983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983110983145983148983141 983108983145983154983141983139983156983151983154983161

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983124983141983160983156 983142983145983148983141

983154983141983137983140983159983154983145983156983141

983123983141983154983145983137983148

15 983116 983106983145 983156 983109 983145 983156 983148 983123 983156

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3539

35

15 983116 983106983145983151983154983141983137983139983156983151983154 983109983160983152983141983154983145983149983141983150983156983137983148 983123983141983156983085983157983152

983122 983148 983124983145 983107 983156 983148983148 983105 983148983145 983156983145

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3639

36

983122983141983137983148 983124983145983149983141 983107983151983150983156983154983151983148983148983141983154 983105983152983152983148983145983139983137983156983145983151983150

983122983137983149983137983150

983120983154983151983138983141

983123983152983141983139983156983154983137983148

983137983139983153983157983145983155983145983156983145983151983150

983155983151983142983156983159983137983154983141

983123983152983141983139983156983154983137983148

983142983145983148983141

983107983137983148983145983138983154983137983156983145983151983150

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141 983141983150983143983145983150983141

983107983151983150983156983154983151983148

983149983151983140983141983148 amp

983154983141983137983148983085983156983145983149983141

983141983150983143983145983150983141

983105983106983106

98309600983160983137

983139983151983150983156983154983151983148

983155983161983155983156983141983149

983125983123983106 983124983107983120983113983120

983105983106983106 983160983120983105983124 983108983137983156983137

983117983137983150983137983143983141983154

983119983120983107983119983120983107

983119983120983107

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3739

37

983109983160983152983141983154983145983149983141983150983156983137983148 983122983141983155983157983148983156983155

bull Communication and controller configuration for real timeapplication complete

bull Plan to run MPC and PID control on both the 3 L and 15 Lbioreactor setups with the Raman probe for real time monitoring ofglucose concentration

bullIn the next few weeks preliminary experimental data for the closedloop feedback control fed-batch runs will be gathered

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3839

38

983107983151983150983139983148983157983155983145983151983150983155 983137983150983140 983110983157983156983157983154983141 983127983151983154983147

bull Fed-batch model has been incorporated into the simulation ofPID and MPC controller technologies

bull MPC performed best for set-point tracking both in thepresence of noise and disturbances from the simulation study

bull Configured and simulated control strategies are to be appliedin real time with online and at-line PAT on two differentbioreactor scales

bull Real time results are to be compared to the results gatheredon a simulation basis

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland

7232019 Mon 1000 Comparison of Controller Technologies for Fed Batch Fermentation S Craven

httpslidepdfcomreaderfullmon-1000-comparison-of-controller-technologies-for-fed-batch-fermentation 3939

39

983105983139983147983150983151983159983148983141983140983143983141983149983141983150983156983155

Prof Brian Glennon

Dr Jessica Whelan

PAT-Bio Research Team

Enterprise Ireland